Vicki Boykis explores the GGUF format used by llama.cpp, which fixed design flaws in GGML and is now the default format. The post also discusses PyTorch models and their traditional persistence using Python pickle.

1m read timeFrom simonwillison.net
Post cover image

Sort: